基于DeepLabV3与GAN的雷达时频混叠多信号智能检测与分离  

Intelligent Detection and Separation of Radar Time-Frequency Aliasing Multi-signal Based on DeepLabV3 and GAN

作　　者：韩文草 孙闽红[1] 王之腾 仇兆炀 HAN Wencao;SUN Minhong;WANG Zhiteng;QIU Zhaoyang(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou,Zhejiang 310018,China;School of Communication Engineering,Army Engineering University of PLA,Nanjing,Jiangsu 210001,China)

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018 [2]陆军工程大学通信工程学院,江苏南京210001

出　　处：《信号处理》2022年第5期1065-1074,共10页Journal of Signal Processing

基　　金：国家自然科学基金(61901149);国防特色学科发展项目(JCKY2019415D002)。

摘　　要：针对当前雷达信号智能检测研究中存在信号数量受限、样式单一、不便于后续处理等问题,本文提出了一种基于DeepLabV3与GAN的雷达时频混叠多信号智能检测与分离方法。首先对接收信号的时域数据通过时频变换得到二维的时频数据,利用DeepLabV3对多信号重叠的时频数据进行检测并且实现信号的分离,对于信号的重叠部分利用GAN对时域信号进行估计重构。实验结果表明,该方法在信噪比(SNR)为-3 dB时,检测平均交并比(mIoU)能够达到了85%以上,在SNR为6 dB时,分离后的信号与原信号相关系数高于0.85。In view of the current research on intelligent detection of radar signals,there are still some problems that have not been resolved,such as the limitation on the number of the signals,the rarity of modulation types and the inconvenience for subsequent processing,this paper proposes a radar time-frequency aliasing multi-signal intelligent detection and separation method based on DeepLabV3 and GAN. First,the time-domain data of the received signal is transformed to obtain the two-dimensional time-frequency data,and DeepLabV3 is used to detect the time-frequency data of multiple signals and realize the signal separation. For the overlapping part of the signal,this paper use GAN to perform the reconstruction of time-domain signal. Results of the experiment show that the mean intersection over union(mIoU)of the detection can reach more than 85% when the SNR is-3 dB,and the correlation coefficient between the separated signal and the original signal is higher than 0. 85 when the SNR is 6 dB.

关 键 词：智能侦察 时频重叠 信号检测 信号分离 深度学习 

分 类 号：TN957.51[电子电信—信号与信息处理]